Chapter 7
Logic Programming and Artificial
Neural Networks
Sebastian Bader,
1 Pascal Hitzler,
2 and Anthony Seda
3
7.1 Introduction
One of the ultimate goals of artificial intelligence is the creation of agents
with human-like intelligence, and many, varied approaches have been made
in attempts to realize this goal. Of course, an agent endowed with humanlike intelligence should be able to represent and reason with well-structured
data and processes, such as those encountered in logic or in mathematics and
related subjects, just as human beings can. On the other hand, that same
agent should also be able to represent and reason with uncertain, noisy, and
incomplete data, again, just as human beings can, at least to a certain extent.
Furthermore, the agent should be able to learn by example and refine the
reasoning process as a result.
These two aspects of the general process of reasoning and intelligence just
considered are complementary and yet are integrated in human intelligence.
Thus, their integration within a single artificial computing system is an important objective in the search for true artificial intelligence.
4 Logic-based
symbolic systems are good implementations of the first, the formal, style of
reasoning, whereas neural networks or connectionist systems are good implementations of the second, less formal, style. They are therefore good candidates, and indeed are among the most prominent such candidates, for attempting this integration, with each representing one of the two aspects. Certainly,
there has been a considerable amount of interest in recent years in exactly this
1 MMIS, Department of Computer Science, University of Rostock, Germany.
2 Kno.e.sis Center for Knowledge-Enabled Computing, Wright State University, Dayton,
Ohio, USA.
3 Department of Mathematics, University College Cork, Cork, Ireland.
4 See [Hitzler and K¨ uhnberger, 2009] for a more detailed discussion of this point.
185
Logic Programming and Artificial
Neural Networks
Sebastian Bader,
1 Pascal Hitzler,
2 and Anthony Seda
3
7.1 Introduction
One of the ultimate goals of artificial intelligence is the creation of agents
with human-like intelligence, and many, varied approaches have been made
in attempts to realize this goal. Of course, an agent endowed with humanlike intelligence should be able to represent and reason with well-structured
data and processes, such as those encountered in logic or in mathematics and
related subjects, just as human beings can. On the other hand, that same
agent should also be able to represent and reason with uncertain, noisy, and
incomplete data, again, just as human beings can, at least to a certain extent.
Furthermore, the agent should be able to learn by example and refine the
reasoning process as a result.
These two aspects of the general process of reasoning and intelligence just
considered are complementary and yet are integrated in human intelligence.
Thus, their integration within a single artificial computing system is an important objective in the search for true artificial intelligence.
4 Logic-based
symbolic systems are good implementations of the first, the formal, style of
reasoning, whereas neural networks or connectionist systems are good implementations of the second, less formal, style. They are therefore good candidates, and indeed are among the most prominent such candidates, for attempting this integration, with each representing one of the two aspects. Certainly,
there has been a considerable amount of interest in recent years in exactly this
1 MMIS, Department of Computer Science, University of Rostock, Germany.
2 Kno.e.sis Center for Knowledge-Enabled Computing, Wright State University, Dayton,
Ohio, USA.
3 Department of Mathematics, University College Cork, Cork, Ireland.
4 See [Hitzler and K¨ uhnberger, 2009] for a more detailed discussion of this point.
185